Behaviorally Explicit Habitat Selection by the Critically Endangered Great Indian Bustard

Home , Great Indian bustard, Naliya, Saara hardwickii

Vol. 45: 55–69, 2021 ENDANGERED SPECIES RESEARCH Published May 27 https://doi.org/10.3354/esr01126 Endang Species Res

OPEN ACCESS

Devil is in the detail: behaviorally explicit habitat selection by the Critically Endangered great Indian bustard

Sutirtha Dutta*, Yadvendradev Jhala

Department of Animal Ecology Conservation Biology, Wildlife Institute of India, Chandrabani PO Box 18, Dehradun 248001, India

ABSTRACT: Habitat management to accommodate ecological needs of threatened species can help abate biodiversity decline. Some species require contrasting microhabitats for different functions, and may prefer patches with ample, diverse microhabitats. We examined this problem for the Critically Endangered great Indian bustard Ardeotis nigriceps in 175 km2 breeding habitat in Kachchh, India. We developed behaviorally explicit resource selection functions (RSFs) by comparing used vs. available microhabitats using binomial generalized linear models that tested hypothesized habitat responses in an information theoretic framework. We identified suitable resource units based on fitted RSF values. We examined if availability of complementary resource units influenced density/usage at the patch level, using line transect distance sampling. Birds preferred agro-vegetation mixture, grassland, high fruit abundance and intermediate grazing density, and they avoided Prosopis thickets for foraging. They preferred moderately tall sward for day resting but shorter sward and less Prosopis for roosting. Nesting females preferred grasslands with relatively tall sward and abundant insects, while displaying males preferred grasslands with shorter sward, far from settlements. Thus, microhabitat selection differed between behaviors and differed from habitat availability. The RSF without behavioral segregation failed to capture these nuances and was non-informative for habitat management. Density/usage at the patch level was correlated with the availability of complementary microhabitats. Thus, protected area management to accommodate diverse life-history requirements may reduce species’ movements over large hostile landscapes and associated mortality. Overall, species requiring complementary microhabitats will benefit from management that promotes habitat heterogeneity. However, habitat use analysis based on behaviorally inexplicit occurrence cannot capture the habitat quality of such species.

KEY WORDS: Ardeotis nigriceps · Conservation · Grasslands · Habitat restoration · Behavior · Foraging · Resource selection function · Distance sampling

1. INTRODUCTION ma nipulation can abate these trends. Hence, understanding resource selection and supplementing pre- Fourteen percent of the world’s birds are threatened ferred resources in breeding areas have benefited (IUCN 2018), with particularly steep declines among many birds. For instance, a study highlighting the im- Indo-Malayan species (Butchart et al. 2004). Reduced portance of tall marsh vegetation and late mowing for fitness in altered habitats has caused much of these corncrake Crex crex (Green 1996) recommended ac- de clines (Tilman et al. 1994), but corrective habitat tions such as delayed mowing that facilitated the re-

© The authors 2021. Open Access under Creative Commons by *Corresponding author: [email protected] Attribution Licence. Use, distribution and reproduction are un - restricted. Authors and original publication must be credited. Publisher: Inter-Research · www.int-res.com 56 Endang Species Res 45: 55–69, 2021

covery of this species (O’Brien et al. 2006). However, vation applications (Boyce & McDonald 1999). These information on habitat use is lacking for many threat- statistical models can explain species−habitat rela- ened species. This is particularly concerning for grass- tionships and predict where species are distributed, lands that are rapidly changing and require urgent and they are frequently applied to manage habitats conservation interventions (White et al. 2000). Bird of endangered species (Rushton et al. 2004). Despite populations are declining faster in grasslands than the growth of resource selection analysis, even in- other biomes of many regions owing to habitat loss, af- volving machine learning approaches (Shoemaker et forestation, mismanagement, land-use intensification al. 2018), their application is not behaviorally expli- and fragmentation (Hovick et al. 2014, SoIB 2020). cit. Their inferences can be biased towards more de - Grassland birds may select complementary habitats tectable behaviors, ignore resource requirements for to meet their diverse ecological needs (Law & Dick - elusive behaviors and run the risk of recommending man 1998). Habitat selection differs between nesting measures that do not encompass the diversity of eco- and feeding usage for short-toed larks Calandrella logical needs of a species. Rahmani (1989) qualitati - brachydactyla (Serrano & Astrain 2005) and black vely described different microhabitats used by great kites Milvus migrans (Sergio et al. 2003). Hence, habi- Indian bustards. However, a comprehensive assess- tat choice should be assessed across multiple behav- ment of the species’ resource selection across behaviors and scales for comprehensive understanding and iors, and the interplay of space-use decisions across holistic management. Species depend on landscape scales, is lacking. These insights will directly help in complementation when their abundance at larger designing and managing their breeding habitats. scales is constrained by the availability of comple- Here, we examined habitat selection by great In- mentary resources at smaller scales (Dunning et al. dian bustards in a semiarid multiple-use landscape, 1992). For these species, availability of complemen- to demonstrate how to manage breeding reserves. tary resources can enable higher usage of an area, by Specifically, we asked: (1) Do these birds select dif- reducing energetic costs of movements, diminishing ferent habitats at fine-grained scales (hereafter predation risk and attracting birds (Choquenot & Rus- referred to as microhabitat or resource units) to fulfil coe 2003). Alternatively, usage by birds may be con- their daily ecological needs, and what are the impli- centrated if complementary resources are available in cations for resource selection studies that ignore proximity, but diffused over a larger area if comple- these behavioral differences? (2) Does density/usage mentary resources are spatially disjoint. Bustards (number of birds per unit area) at larger grain sizes (family Otididae), a group of globally threatened depend on the availability of complementary grassland birds, offer an ideal system to test this pos- resource units at finer scales such that strategic inser- tulate because of their requirement for heterogeneous tion of missing resource units can promote greater habitats and growing de pendence on conservation patch usage? To this end, we collected microhabitat efforts (Collar et al. 2017). data from 100 m radius plots at locations used by The great Indian bustard Ardeotis nigriceps is Criti- great Indian bustards for foraging, day resting, night cally Endangered with around 100−150 birds left in In- roosting, courting and nesting, along with available dia and Pakistan (Dutta et al. 2011, BirdLife Inter- (random) locations. We developed resource selection national 2018). Its habitats, mainly arid−semiarid functions (RSFs) for each behavior and a common grass lands, are marginalized as ‘unproductive waste- RSF for all behaviors by testing ecological predic- lands’ and are experiencing infrastructural develop- tions with used vs. available habitat data using bino- ment and intensive land uses (Dutta 2018). The Indian mial generalized linear models (GLMs) in an infor- government is implementing recovery actions for this mation-theoretic framework (Manly et al. 2002). We species that include restoration of breeding habitats compared inferences be tween behaviorally explicit (Dutta et al. 2013). Enclosures are being established in and common RSFs, to answer the first question. For Thar (Rajasthan), Kachchh (Gujarat) and other range the second question, we classified random locations areas, to reduce anthropogenic disturbances, increase as ‘suitable’ or ‘unsuitable’ for a behavior based on herbaceous cover and improve breeding success. fitted RSF values, and computed microhabitat diver- Since the species faces high mortality risk due to sity at a larger scale (transects) from the frequency of power-line collisions, habitat management to increase ‘suitable’ locations for various behaviors. Finally, we bird density/usage in infrastructure-free protected en- examined if the species’ density in transects, esti- closures can aid in the recovery of this species. mated using distance sampling (Burnham et al. Inferences on habitat suitability by comparing used 1980), depended on microhabitat diversity, using vs. available locations have a long history in conser- GLM. Dutta & Jhala: Habitat selection by great Indian bustards 57

2. MATERIALS AND METHODS (Dutta 2012). The regional climate is semiarid with high temperature variations (0−5°C in January to 2.1. Study area 40−45°C in May) and scant rainfall (mean 384 mm, range 78−888 mm during 2000−2010). Tropical thorn We conducted fieldwork from 2007 through 2011 in forest/scrub and grasslands (Champion & Seth 1968) Abdasa tehsil (precinct) of Kachchh, Gujarat, India. are grazed by free-ranging livestock, and are inter- This landscape harbors a small and declining popu- spersed with Sorghum bicolor, Pennisetum glaucum, lation of great Indian bustards that uses a central Arachis hypogaea and Gossypium spp. cultivation contiguous patch of 175 km2 during the breeding sea - (Fig. 1). Recently, cultivation has expanded with son (summer: March−June, monsoon: July− October) more frequent tilling and year-round cropping, lead-

Fig. 1. Study area and sampling design. (a) Prime breeding habitat of great Indian bustard (GIB), showing major land covers. (b) Location of study area in Abdasa, Kachchh, India. (c) Sampling scheme for measuring bird usage and habitat variables. (d) Distribution of random locations. (e) Locations used by GIB for foraging, resting, roosting, courting and nesting behaviors during 2007−2011 58 Endang Species Res 45: 55–69, 2021

ing to higher livestock dependence on remnant pas- sampled the same variables at 2−3 random locations tures — a scenario that typifies the range of great in each 1 km2 grid overlaid on the study area. Ran- Indian bustards (Dutta & Jhala 2014). The Lala Bus- dom locations were generated in ArcGIS, and 1 loca- tard Sanctuary (2 km2) and 2 enclosures near Naliya tion was sampled per grid in a season. However, a (20 km2) and Kunathia (4 km2) villages have been small fraction (<10%) of grids could not be sampled established here for bustard conservation. representatively and adequately due to logistic constraints. These data represented the general habitat, not the unused area, and was preferred because the 2.2. Data collection and analyses rarity of birds would contaminate unused locations with pseudo-absences (Johnson et al. 2006). 2.2.1. Bird-habitat surveys

We searched the study area on foot and on existing 2.2.2. Fine-grained RSFs trails with a motorcycle (driver with a pillion observer, traveling at <10 km h−1) from dawn through dusk with We described the habitat of foraging (n = 76 loca- a break during the day (12:30−14:30 h, when bird actions), resting (24), roosting (24), courting (10) and tivity ceased). We covered the area 6 to 10 times dur- nesting (12) locations, using box and whisker plots ing the breeding season over 2 years (2008−2009). We and descriptive statistics for straightforward inter- obser ved bird behavior using 8 × 42 binoculars from pretation of the species’ requirements. Before further suitable vantage points (at 200−500 m distance from analysis, we transformed the data as follows. (1) Veg- the focal birds) that allowed an unobstructed view etation variables were z-standardized using their without altering their natural behavior. We recorded seasonal means and standard deviations, to remove landmarks and projected locations with a GPS unit rainfall-induced difference in their summer and corresponding to the following behaviors: (1) foraging, mon soon measurements and the ensuing noise in where birds were intensively feeding or searching for habitat selection inferences that were not relevant to food; (2) resting, where birds retired by sitting with this study. (2) Synthetic factors were extracted from the head drawn against the body for at least 20− 30 strongly correlated habitat variables using factor min in the daytime; (3) roosting, where birds retired ana lysis (following Graham 2003), to reduce data after sunset and ample fecal/feather deposits were re- dimensionality and to avoid multicollinearity. covered during the following morning indicating that Thereafter, we examined habitat selection from birds had stayed at that spot overnight; (4) courtship, used and available locations (design I of Manly et al. where males performed sexual displays from arenas, 2002) using GLMs with logit link and binomial errors including territorial rituals; and (5) nesting, where fe- (McCullagh & Nelder 1989). Since used and avail- males in cubated eggs and egg-shells were subse- able locations were sampled independently without quently re trie ved. To increase sample size, we used replacement, the RSF was proportional to the expo- nest information from surveys conducted by the State nential function:

Forest Department around the same years as our ββ+++xx.... β e 011 kk (1) study, and these data might represent nests of the β same females. Similarly, courtship locations were ob- where xp = 1…k are habitat variables and p = 1…k tained from different display arenas of the same are model parameters to be estimated (Manly et al. males and over multiple years. 2002). Here, GLM estimates the probability of a We visited ‘used’ locations after bird(s) moved resource unit being used if it is sampled, as: away, and measured variables that could potentially ’ xx.... eββ011+++ βkk influence habitat selection. Since most birds distin- (2) ββ’011++x .... + βkkx guish habitat based on structural characteristics (Co- 1+ e β β dy 1985), we measured 10 variables depicting land where the intercept ’0 is modified from 0 to include cover and vegetation structure, 2 variables for food sampling probabilities of used and available units. availability and 3 variables for human disturbances. First, we modeled RSFs for ‘general occurrence’ by Descriptions, measurements and postulated effects pooling used locations across all behaviors, to mimic of these variables are provided in Table 1. We meas- the common approach of behaviorally simplistic ha - ured these variables in nested circular plots of 100 m bi tat analysis. Since behavioral differences pre- radius (9 variables) and 10 m radius (6 variables). To cluded hypotheses that were generalizable across characterize ‘available’ habitat, we systematically be haviors, we used exploratory (all subsets) model Dutta & Jhala: Habitat selection by great Indian bustards 59

Table 1. Collection of habitat variables from locations used by great Indian bustards and random locations along with their a priori predicted effects on resource selection. Variables were measured at different scales (coarse: 100 m and fine: 10 m radii) to strike a balance between (i) the mobility of birds: great Indian bustards can cover 1900 m h−1 (S. Dutta unpubl. data); and Gray et al. (2007) used 50 m radius plots for the related, but smaller, Bengal florican; (ii) inherent scale disparity among variables: land cover is homogeneous and therefore measurable at a larger grain size than vegetation structure; and (iii) practicality of sampling: fruiting intensity can be assessed over larger area than insect abundance. Sources: (1) Johnsgard (1991), (2) Wolff et al. (2001), (3) Rahmani (1989), (4) S. Dutta pers. obs., (5) Magana et al. (2010), (6) Stephens & Krebs (1986), (7) Dutta (2012), (8) Hilbert et al. (1981), (9) Lavee (1988), (10) Osborne et al. (2001), (11) Dutta et al. (2011)

Feature Variables Measurement and processing A priori hypotheses and predictions

Land cover Grassland We recorded the dominant land cover Bustards evolved in open habitats, and Scrubland in 100 m radius of a location by ocular would generally prefer grassland (1); but Agriculture assessment. Co-dominance of grass/ moderately human-altered land uses offer Agro-veg mix scrub and agriculture was recorded as higher food diversity (2) and can be ‘agro-veg mix’ selected for foraging Vegetation Grass height We recorded vegetation characteristics Because of contrasting preferences of characteristics Grass cover on a 5-point scale by classifying their concealment and visibility for different (coarse grain) Shrub height range at (a) 5, 20, 40, 70, 110 cm (grass ecological needs, sparse vegetation will be Shrub cover height); (b) 10, 20, 40, 60, 100% (grass selected for roosting to enhance surveil- cover); (c) 0.5, 1.0, 2.0, 3.0, 5.0 m (shrub lance and reduce ambush predation, and height); and (d) 5, 15, 25, 40, 70% (shrub for courtship to allow transmission of cover), based on ocular assessment in a sexual advertisement; moderately tall, 100 m radius around a location dense vegetation for resting to reduce thermal stress and for nesting to provide Vegetation Bare ground % We measured vegetation stratification in concealment; and intermediate vegetation stratification Vegetation a 10 m radius around a location as the to optimize food abundance and detection, (fine grain) (<25 cm) % proportion of 10 vertical hits of a 1 m for foraging (3, 4, 5). Tall, thorny shrubs Vegetation calibrated pole where the ground may be avoided because of hindrance to (25−50 cm) % vegetation cover crossed 25, 50 and visibility and movements (3, 4) Vegetation 100 cm marks (50−100 cm) % Vegetation (>1 m) % Resource Insects We counted insects in a 20 × 2 m2 strip While birds may track food resources (6), availability Fruit in 10 m radius around a location and the patchier food type (fruits rather than indexed fruit availability as 0 (low) to insects) can exert greater influence on the 4 (high) based on fruiting shrubs/trees in choice of foraging site (4) a 100 m radius around a location Key shrub Zizyphus We ranked the dominance of these shrubs Z. nummularia, a stunted shrub that great dominance Prosopis in a 100 m radius around a location as 0 Indian bustards frequently feed on (win- (absent) to 3 (most abundant) ter−summer) (3, 4, 7) can attract foraging use, whereas P. juliflora, an invasive exotic shrub that forms dense thickets, obstruct- ing visibility and facilitating predators, will be generally avoided (4) Anthropogenic Grazing We quantified grazing pressure using Following the grazing optimization concept disturbances pressure 3 indices: (1) visible impacts of grazing/ (8), we postulated that intermediate browsing and trampling (MacDonald et al. grazing pressure can favor foraging use by 1998) in a 100 m radius around a location, optimizing vegetation productivity and and counts of (2) pellets/dung and (3) making dung beetles (an important food) tracks along a 20 m transect at a location. transiently available. However, human We scored these indices as 0 (low) to 4 disturbances associated with grazing can (high) and averaged these values into a deter nesting use (9) composite score that was more comprehensive than any single measure Distance to We recorded the distance of a location to While birds may generally avoid human village the nearest village and metal road from artifacts to reduce mortality risks (10, 11), Distance field-digitized maps in ArcGIS 9.2 such a response can be particularly pro- to road nounced during resting/roosting, when birds are stationary and more vulnerable to attacks 60 Endang Species Res 45: 55–69, 2021

building for general occurrence. Next, we developed towards high-usage areas. We modeled perpendicular an RSF for each behavior by building candidate mod- distances of bird sightings from the transect using els corresponding to alternative hypotheses (Table 1). hazard rate and half-normal detection functions, and The global models included variables with greatest a selected the best model using AICc (Akaike 1974) and priori importance, to ensure ≥5 ‘used’ data points per GOF tests in the program DISTANCE v 5.2 (Thomas estimated effect (following Vittinghoff & McCulloch et al. 2010). We estimated a common detection proba- 2007). We examined the goodness of fit (GOF) and bility since visibility did not vary much across the variance inflation factors of the global models (Quinn study area, and applied it to convert encounter rates & Keough 2002). On obtaining satisfactory diagno - at the transect level into density estimates. Distance ses, we compared candidate models in an infor - sampling has been applied to other bustards as well mation theoretic framework (Burnham & Anderson (Tourenq et al. 2005), and violations of its assumptions 2002). We used the average (when competing mod- will not affect the relative space usage, which was the els were <2 ΔAICc units apart, where AICc is parameter of interest. Akaike’s information criterion corrected for small sample size) or the least AICc model in its exponent form as the RSF (Boyce 2006). We interpreted behav- 2.2.4. Resource complementarity and ioral differences in RSFs based on the direction, mag- congruence across scales nitude and precision of model parameters. To evalu- ate the reliability of inferences, we reiterated the RSF Fitted RSF values predicted the relative probability modeling for each behavior over 10 runs with differ- of selecting random locations for a behavior (Manly et ent subsets of available locations (50% locations al. 2002). Subsequently, we classified random lo - chosen at random without replacement in a run). We cations with RSF values exceeding the first quartile used the mean and CV of variable effect sizes across RSF value of used locations (implying relatively high bootstrapped models, to examine the variability chance of selection) as suitable resource units for that among individual RSF runs and for robust prediction behavior. For robust prediction, we averaged the lo- of microhabitat suitability. Despite subsampling the cation’s predicted state (suitable/unsuitable) across data, sampling ratios were large (5−10 available lo- 10 bootstrapped RSFs (see Section 2.2.2 and Tables S1 cations per used location) for most behaviors, which to S3 in Sup plement 2 at www. int-res. com/ articles/ has been shown to reduce inference bias (Baasch et suppl/ n045 p055 _supp/ ). We also identified resource al. 2010, Nad’o & Kaňuch 2018). We conducted these units for general occurrence in a similar way that re- analyses in R v 2.13.0 (R Core Team 2013) (R script flected simplistic (behaviorally inexplicit) inferences data code is available in Supplement 1 at www. int- on habitat selection. res. com/ articles/ suppl/ n045 p055 _ supp/). Next, we examined if resource characteristics, and hence, suitability of random locations, were similar across behaviors and general occurrence. We tested 2.2.3. Coarse-grained space use this postulate using scatter plots and Pearson’s correlation analysis (Quinn & Keough 2002) on pairs of We estimated great Indian bustard density (number predicted RSF values (relative suitability of locations) per area) along line transects using distance sampling for various behaviors and general occurrence. (Burnham et al. 1980), as a measure of their usage at Finally, we examined if availability and complemen- the coarse grain. We surveyed 15 transects of average tarity of resource units influenced usage at the coarse 4 ± 0.7 (SD) km length and 1 km width on either side, grain. For this, we quantified ‘resource availability’ on a slow-moving motorcycle with a pillion observer (RA) in ~8 km2 transect areas, as the proportion of ran- during the prime bird activity period (07:00−11:00 and dom locations that were suitable for any behavior 17:00−19:30 h, Rahmani 1989). We resurveyed each (hereafter, useable resource units). We quantified ‘re- transect 4 times during the breeding season for 3 source complementarity’ (RC) as the product of the years (2008−2010), resulting in an average of 12 ± 2 frequency of useable resource units in the transect temporal replicates per transect. This grain size (aver- area and the exponent form of the Shannon-Wiener age 8 km2) offered varying de grees of structural com- index, computed from proportions of suitable resource plexity that could influence habitat usage, and units for various behaviors. This index served as a sur- matched the daily range of this species (2.6 ± 5.0 km2 rogate for the effective number of complementary re- based on 1 tagged bird, authors’ un publ. data). Tran- sources (Jost 2006) and obtained higher values for sects were placed on available dirt trails with no bias transect areas with greater availability of diverse re- Dutta & Jhala: Habitat selection by great Indian bustards 61

source units. We modeled great Indian bustard density vegetation mixture > agriculture. Birds used grass- at transects using Gaussian GLMs to examine the ef- lands for nesting and courtship, and grassland or fects of RA and RC against the null model (constant agro-vegetation mixture for foraging; whereas, use of density). We used AICc and R2 statistics to compare agricultural land was minimal. Available vegetation these models (Quinn & Keough 2002). was structurally dominated by bare ground ≈ short (<25 cm) sward > moderate (25−50 cm) sward > tall (50−100 cm) sward. Birds used relatively tall sward 3. RESULTS for nesting, relatively short sward for courtship and roosting, and moderate sward for resting. 3.1. Microhabitat characteristics Since vegetation variables were collinear (|ρ| > 0.4, see Table S5 in Supplement 3), they were summarized Great Indian bustard microhabitats differed be - into 3 synthetic factors that cumulatively ex plained tween behaviors and from available locations in sev- 64% variance in data. These factors represented gra- eral ways (Fig. 2; Table S4 in Supplement 3 at www. dients of: (1) openness to moderately tall and dense int-res. com/ articles/ suppl/ n045 p055 _ supp/). Avail- sward (hereafter, sward biomass factor), (2) short to able land cover comprised grassland > scrub ≈ agro- tall herbaceous cover (hereafter, short vs. tall sward

5 a 5 b 4 c 6 d 4 4 3 5 3 3 2 4 2 2 1 3

1 1 Shrubbiness Sward biomass

0 Short vs. tall sward 0 0 2 Moderate vs. tall sward 4 4 4 4 e f g h

3 3 3 3

2 2 2 2 dominance

1 1 1 1 Grazing intensity Distance to road Distance to settlement

0 0 0 Prosopis 0 A B D F Rd Rn 40 4 4 i j k 1.0 l

30 3 3 0.8

0.6 20 2 2

dominance 0.4 10 1 1 0.2 Insect availability Fruit availability

0 0 0 Frequency occurrence

Zizyphus 0.0 ABDFRd Rn A B D F Rd Rn A B D F Rd Rn A B D F Rd Rn Scrub Agro-veg Grassland Agri

Fig. 2. Quartile distribution of microhabitat variables for great Indian bustards in Kachchh, India: (a) sward biomass; (b) short vs. tall sward; (c) shrubbiness (factors); (d) moderate vs. tall sward; (e) grazing intensity index; distances (km) to (f) nearest settlement and (g) paved road; (h) invasive Prosopis dominance, (i) insect count per 10 m, (j) fruit availability, (k) native Zizyphus dominance, and (l) relative frequency (proportion) of land covers. ‘Available’ habitat is marked with A (open boxplots). ‘Used’ locations (solid boxplots) are as follows: B: breeding/nesting, D: courtship/display, F: foraging, Rd: day resting and Rn: night roosting. Notch represents median, box represents central 25–75% observations, whiskers represent extreme lower (2.5– 25%) and upper (75–97.5 %) observations. Thickness of boxplots reflects sample size, and non-overlapping notches indicate difference in medians of a variable 62 Endang Species Res 45: 55–69, 2021

factor) and (3) shrub biomass (hereafter, shrubbiness 3.3. Usage at coarse resolution factor), respectively (Table S6 in Supplement 3). Transect surveys of 917 km yielded 55 detections of great Indian bustards along 15 spatial replicates dur- 3.2. Behaviorally explicit RSFs ing the breeding season. A half-normal detection function with cosine adjustment best fitted the dis- Comparison between alternative hypotheses on for- tance data (lowest AICc and GOF-χ2 = 0.60, df = 3, aging use showed that birds selected agro-vegetation p = 0.90). Effective strip width was estimated to be mixture and grassland over agriculture, higher fruit 231 m (95% CI: 177−301 m). Bustard density was abundance, Zizyphus (stunted fruiting shrubs), inter- estimated to be 0.17 (0.13−0.23) ind. km−2, indicating mediate grazing intensity and relative proxi mity to set- that 29 (22−39) birds used the area. Bird density var- tlements, but avoided Prosopis thickets (Tables 2 & 3). ied substantially across transects (CV 102%), rang- Multi-model inference based on the top 3 hy po theses ing between 0 and 0.65 ind. km−2. Density across on day resting showed that birds selected relatively transects (n = 12; 3 transects dropped due to inade- tall−dense sward while other variables lacked precise quate habitat data) was best explained by the avail- effects. However, we omitted resting RSFs from further ability of complementary resources (RC: Akaike analysis since the full model showed a poor fit to the weight W = 0.65, R2 = 0.43), showing a positive rela- data (Nagelkerke R2 = 0.05). Multi-model inference tionship be tween the 2 variables (β = 0.74 ± 0.24 SE) based on the top 3 hypo theses for night roosting (Fig. 3; Table S7 in Supplement 3). Residual plot showed that birds selected shorter sward and avoided diagnosis indicated acceptable fit of this model to the Prosopis. For nesting use, the full model received maxi- data (Fig. S2 in Supplement 3). mum support and indicated selection for grassland, tall sward and higher insect abundance. For courtship use, the full model ob tained maxi mum support and indi- 4. DISCUSSION cated selection for grassland, shorter sward and greater distance from settlements. Model bootstrapping Bustards are declining globally, with nearly two- with different subsets of random locations indicated thirds of the taxa listed as Threatened or Near that the above inferences were reasonably robust, Threatened (BirdLife International 2018). Their con- since the effect sizes of important variables were esti- servation is a global challenge and necessitates com- mated with <20% CV across model runs for foraging, prehensive understanding of habitat requirements. roosting, courting and nesting behaviors (Table 3). The great Indian bustard population in Kachchh has As evident from RSF model rankings and parame- declined since the study to a current total of about ter estimates, factors influencing habitat selection 5 birds (D. Gadhvi unpubl. data). Their decline is and their effects varied between behaviors. Re source largely attributed to collision with overhead power- selection values of locations were weakly correlated lines and habitat alteration. Although biotelemetry between types of use (r ranged from −0.04 [roosting × can provide robust insights into resource selection, nesting] through 0.18 [foraging × roosting]; Fig. S1 in research funding and permit restrictions have often Supplement 3), indicating that a single microhabitat precluded tagging of a large sample of individuals type is not suitable for all behaviors. Although 52 ± that is necessary for such inferences. This makes 5% (95% CI) of available locations were found to be direct observations at the population level the most useable, resource units for foraging were more pre- feasible available approach for many species. valent (37 ± 5%) than for roosting (26 ± 5%), court - ship (20 ± 4%) and nesting (3 ± 2%). Our exploratory model building for RSFs of general 4.1. Effects of habitat variables occurrence (14 top models within <2 ΔAICc units) in - dicated that grassland and agro-vegetation mixture Great Indian bustards selected grassland and were selected over scrub and agriculture, Prosopis avoi ded agricultural land. Selectivity of grasslands was avoided, and short sward was preferred over tall was particularly strong for courtship and nesting ac- sward (Table 3). Effects of other habitat variables tivities. Bustards prefer flat, undisturbed grasslands were imprecise. Resource selection values of locations for sexual advertisement and nesting (Osborne et al. for general occurrence were incongruent with those 2001), and this postulate was corroborated by our of specific behaviors (r ranged from −0.06 [use × roost- finding. India’s grasslands are facing agricultural ex - ing] through 0.35 [use × foraging]; Fig. S1). pansion and intensification, involving reduced fallow Dutta & Jhala: Habitat selection by great Indian bustards 63

Table 2. Ranking of alternative hypotheses on (a) foraging, (b) resting, (c) roosting, (d) nesting and (e) courtship resource selections of great Indian bustards in Kachchh, India. Binomial generalized linear models fitted on variables: land cover (hab) or grassland (hab-grl) vs. other land covers, sward biomass (swd-bio), tall vs. short sward (tl-swd) and shrubbiness (shrb) factors, insect (ins) and fruit (frt) availability, Prosopis (pro) and Zizyphus (ziz) dominance, grazing intensity (grz), distance to settlement (dst-set) and distance to road (dst-rd). Quadratic effects of some variables (e.g. sward biomass swd-bio2 and grazing intensity grz2) were included in models to test predictions that immediate values of these variables might be selected. Sum- mary statistics include Akaike weight (W), Akaike’s information criterion corrected for small sample size (AICc), difference of AICc from the lowest AICc value (ΔAICc), deviance (−2logL) and model parameters (K)

W ΔAICc AICc −2logL K

(a) Candidate models for foraging use hab + pro + ziz + frt + grz + grz2 + dst-set 0.38 0.00 279.76 238.86 10 hab + shrb + ziz + frt + grz + grz2 + dst-set 0.14 2.00 281.76 240.86 10 hab + pro + ziz + frt + grz + grz2 0.14 2.02 281.78 245.05 9 hab + swd-bio + swd-bio2 + pro + ziz + frt + grz + grz2 + dst-set 0.07 3.37 283.13 233.84 12 hab + ziz + frt + grz + grz2 0.05 4.14 283.90 251.32 8 hab + swd-bio + swd-bio2 + pro + ziz + frt + dst-set 0.04 4.32 284.08 243.18 10 hab + shrb + ziz + frt + grz + grz2 0.04 4.46 284.22 247.49 9 hab + swd-bio + swd-bio2 + pro + ziz + frt + grz + grz2 0.03 5.06 284.82 239.74 11 hab + swd-bio + swd-bio2 + shrb + ziz + frt + grz + grz2 + dst-set 0.03 5.31 285.08 235.79 12 hab + ziz + frt 0.02 6.13 285.89 261.55 6 hab + swd-bio + swd-bio2 + shrb + ziz + frt 0.00 8.94 288.70 251.97 9 hab + swd-bio + swd-bio2 + tl-swd + shrb + pro + ziz + frt + ins + 0.00 9.74 289.50 223.22 16 grz + grz2 + dst-set + dst-rd swd-bio + swd-bio2 + pro + ziz + frt 0.00 10.17 289.94 265.60 6 hab + swd-bio + swd-bio2 + pro 0.00 13.35 293.11 264.66 7 hab 0.00 16.08 295.84 279.68 4 ziz + frt + ins 0.00 19.22 298.98 282.82 4 frt 0.00 25.52 305.29 297.24 2 pro 0.00 25.88 305.64 297.60 2 grz + grz2 0.00 29.83 309.60 297.50 3 swd-bio + swd-bio2 0.00 30.79 310.55 298.45 3 dst-set 0.00 31.01 310.77 302.73 2 null 0.00 33.65 313.41 309.39 1 ins 0.00 34.78 314.54 306.50 2 swd-bio 0.00 35.18 314.94 306.90 2 grz 0.00 35.36 315.12 307.07 2 (b) Candidate models for resting use swd-bio 0.31 0.00 147.47 139.41 2 swd-bio + dst-set 0.16 1.29 148.76 136.64 3 swd-bio + shrb 0.16 1.37 148.83 136.71 3 null 0.10 2.34 149.80 145.78 1 swd-bio + shrb + dst-set 0.08 2.63 150.10 133.90 4 shrb 0.06 3.41 150.87 142.81 2 dst-set 0.05 3.54 151.01 142.95 2 (c) Candidate models for roosting use swd-bio + tl-swd + pro 0.32 0 129.47 113.27 4 tl-swd + pro 0.31 0.09 129.57 117.45 3 swd-bio + tl-swd + pro + dst-set 0.29 0.21 129.69 109.38 5 hab + swd-bio + tl-swd + pro 0.05 3.69 133.17 104.59 7 hab + swd-bio + tl-swd + pro + dst-set 0.03 4.52 133.99 101.25 8 tl-swd 0 11.82 141.3 133.24 2 swd-bio + tl-swd + dst-set 0 12.07 141.54 125.34 4 pro 0 12.19 141.67 133.61 2 null 0 20.33 149.8 145.78 1 dst-set 0 21.18 150.66 142.6 2 hab 0 21.31 150.78 134.58 4 swd-bio 0 21.82 151.29 143.23 2

Table continued on next page 64 Endang Species Res 45: 55–69, 2021

Table 2 (continued)

W ΔAICc AICc −2logL K

(d) Candidate models for nesting use hab-grl + tl-swd + ins 0.98 0.00 38.49 22.27 4 tl-swd + ins 0.02 7.90 46.39 34.26 3 hab-grl + ins 0.00 13.99 52.48 40.36 3 hab-grl + tl-swd 0.00 24.47 62.96 50.83 3 ins 0.00 27.68 66.17 58.10 2 tl-swd 0.00 38.84 77.33 69.27 2 hab-grl 0.00 39.32 77.81 69.75 2 null 0.00 53.31 91.80 87.78 1 (e) Candidate models for courtship use hab-grl + tl-swd + dst-set 0.71 0.00 67.02 50.80 4 hab-grl + dst-set 0.19 2.67 69.69 57.56 3 hab-grl + tl-swd 0.07 4.66 71.68 59.56 3 hab-grl 0.02 7.35 74.37 66.30 2 tl-swd + dst-set 0.01 7.99 75.01 62.88 3 tl-swd 0.00 10.70 77.72 69.66 2 null 0.00 13.35 80.37 76.35 1

Table 3. Resource selection functions for foraging, resting, roosting, nesting, courtship and general occurrence of great Indian bustards in Kachchh, India: (a) estimated effects β (SE) of habitat variables based on a random subset of available locations, (b) bootstrapped mean (CV) effects of same variables from model iterations with different subsets of available locations. Blank cells indicate variable effects that were either not tested or not selected in the top/average model. Squared terms indicate potential quadratic effects that were tested since birds might select intermediate values of some variables for particular behaviors. ΣW: summed Akaike weight of variables across models explaining general occurrence

(a) Features Variables Foraging Resting Roosting Nesting Courtship Occurrence ΣW

Land cover Agro-vegetation 2.33 (1.09) 0.31 (0.11) 1 Grassland 2.27 (1.09) 3.47 (1.45) 2.40 (0.87) 0.38 (0.11) 1 Scrubland 1.48 (1.14) 0.12 (0.12) 1 Vegetation Sward biomass 0.46 (0.24) −0.40 (0.28) 0.01 (0.02) 0.05 structure Sward biomass2 −0.002 (0.005) 0.05 Tall vs. short sward −0.97 (0.28) 2.33 (0.82) −0.93 (0.47) −0.05 (0.03) 0.95 Shrubbiness −0.18 (0.24) Shrub Prosopis −0.40 (0.23) −1.63 (0.78) −0.07 (0.03) 1 dominance Zizyphus 0.36 (0.15) 0.03 (0.03) 0.76 Food Fruit availability 0.34 (0.14) 0.04 (0.03) 0.81 resources Insect availability 0.22 (0.06) Human Grazing intensity 1.97 (0.91) 0.03 (0.06) 0.35 distur- Grazing intensity2 −0.50 (0.22) −0.02 (0.02) 0.68 bances Dist. to settlement −0.49 (0.24) 0.29 (0.33) −0.51 (0.38) 1.61 (0.67) −0.01 (0.02) 0.41 (b) Features Variables Foraging Roosting Nesting Courtship

Land cover Agro-vegetation 2.41 (0.13) 1 (0.37) Grassland 2.2 (0.15) 0.79 (0.3) 3.03 (0.15) 2.3 (0.09) Scrubland 1.26 (0.26) −0.54 (0.71) Vegetation Sward biomass −0.47 (0.2) structure Sward biomass2 Tall vs. short sward −1.03 (0.12) 2.47 (0.28) −0.84 (0.17) Shrubbiness Shrub Prosopis −0.45 (0.17) −1.71 (0.08) dominance Zizyphus 0.46 (0.11) Food Fruit availability 0.33 (0.19) resources Insect availability 0.11 (0.75) Human Grazing intensity 2.02 (0.18) distur- Grazing intensity2 −0.49 (0.16) bances Dist. to settlement −0.61 (0.15) −0.52 (0.24) 1.37 (0.15) Dutta & Jhala: Habitat selection by great Indian bustards 65

0.6 2008, Gray et al. 2009), while low-intensity grazing )

2 allows taller vegetation that provides concealment 0.5 for nesting (Magana et al. 2010). Nesting females 0.4 might also prefer low-intensity grazing to avoid dis- turbance caused by livestock herds and their guard 0.3 dogs; incubating MacQueen’s houbara bustards 0.2 Chla mydotis macqueenii have been observed to 0.1 temporarily leave nests when disturbed by livestock

Bird densit (ind. km (Koshkin et al. 2016). Although Koshkin et al. (2014, 0.0 2016) found that density and nesting productivity of 0.3 0.4 0.5 0.6 020.2 070.7 C. macqueenii were independent of moderate levels Resource availability x heterogeneity of sheep density in Kyzylkum desert, Uzbekistan, Fig. 3. Relationship between great Indian bustard density these studies did not capture the most extreme live- (re presenting coarse-grained use intensity) and availability stock densities (>80 km−2) that would resemble our of complementary resources along transects in Kachchh, India study area (25 animal units or 100 sheep/goat equiv- alents km−2, see Dutta 2012). Similarly, high livestock periods, mechanized farming, inorganic pesticides, density was considered as the major reason behind fertilizers and groundwater irrigation that are un - poor nesting success of C. macqueenii in Israel favorable for breeding bustards, much like their glo - (Lavee 1988). However, habitat selection for other bal counterparts (Dutta et al. 2013). However, the behaviors was unrelated to livestock grazing, indica- species also selected seasonal croplands interspersed ting that this land use was compatible with non-nest- within grassland/scrub (agro-vegetation mixture) for ing ecological requirements. These findings support foraging. Bustards are generalist feeders (Lane et al. the prescribed action of fencing relatively small 1999), and great Indian bustards feed on insects, (<10 km2) and known breeding sites to regulate fruits, harvested crops and plant matter (Dutta 2012). grazing and improve recruitment (Dutta et al. 2013). Foraging birds might benefit from the structural het- Curbing grazing in relatively smaller areas safe- erogeneity of agro-vegetation mixture that provides a guards non-breeding ecological needs and pastoral higher diversity and/or abundance of food items en- resources. Nevertheless, the availability of better compassing natural and agricultural subsets (Lane et forage inside enclosures compared to the heavily al. 2001). Wolff et al. (2001) reported relatively higher grazed surroundings results in occasional conflicts density of little bustards Tetrax tetrax in mixed zones between pastoralists and wildlife managers. Such of steppe and extensive farms, and proposed that conflicts need to be pacified through participatory their abundance increased as natural grasslands actions (e.g. community fodder farms) or incentives were replaced by extensive agriculture, but de - (e.g. subsidizing market fodder). creased with intensification of farming. Similarly, we Great Indian bustards selected shorter shrub infer that large-scale conversion of grasslands into heights (1.18 ± 0.13 m 95% CI), even more so for intensive agriculture is detrimental, but low-intensity nesting (0.40 ± 0.20 m) and courtship (0.95 ± 0.39 m), seasonal agriculture interspersed within grasslands is compared to available habitat (1.72 ± 0.12 m). Bus- compatible with great Indian bustard conservation. tards depend on long-distance visibility for anti- Hence, regulating the expansion of agriculture and predator vigilance and sexual advertisement, which managing its spatial arrangement can benefit the is compromised in excessively shrubby environ- species, as has been shown for the related lesser flori- ments. We found stronger avoidance of particular can Sypheotides in dicus (Dutta & Jhala 2014). The shrubs such as Prosopis juliflora than shrubbiness Indian government may consider declaring priority per se, as tall−dense Prosopis thickets disrupt mobil- landscapes as eco-sensitive zones for such land-use ity and vision more than other shrubs. Historically, regulations (Dutta et al. 2011). Forest Departments have planted arid−semiarid Great Indian bustards selected relatively less landscapes with shrub/tree species including P. juli- grazed areas for nesting and more grazed areas for flora, Acacia tortilis, A. bivenosa and Glyricidium court ship. Grazing intensity underpins vegetation spp. (Dutta 2018). Our findings suggest that this structure in this landscape (Dutta & Jhala 2014). prac tice is detrimental to great Indian bustards. High-intensity grazing promotes shorter vegetation However, small patches of stunted fruiting shrubs that is preferred by displaying males for transmitting such as Zizyphus can increase food availability, and sexual signals over larger distances (Hingrat et al. foraging birds preferred its presence (Dutta 2012). 66 Endang Species Res 45: 55–69, 2021

Avoidance of shrubby areas with poor visibility and analyses and conservation. Firstly, a common RSF preference of relatively tall sward (<100 cm) for con- based on pooled occurrence data may not identify re- cealment by nesting females corroborated the con- source characteristics that are critical to the species’ clusion of Magana et al. (2010) that nest site selection conservation. Our behaviorally inexplicit model did not was a trade-off between visibility and concealment show any precise effect of vegetation structure, except in the related great bustards Otis tarda. for a slight selection of shorter sward, thus rendering a Insect availability was markedly higher at nesting lopsided view of the species’ re quirements. If this sites. Greater insect abundance can be an indirect model were used to inform habitat management, it effect of preferring less grazed and relatively tall would result in proliferation of short grasslands, sward (Jerrentrup et al. 2014) for nesting. Alterna- leading to depletion of relatively tall swards, which are tively, breeding females might prefer nesting in important for nesting in this landscape. Secondly, infer- areas with greater insect abundance to meet dietary ences from such pooled occurrence data would be bi- requirements during the critical incubation and ased towards easily de tec table behaviors. Resource chick-rearing periods when movement is restricted characteristics of less detectable yet critical behaviors (Bretagnolle et al. 2011). We did not find such strong such as nesting or roosting would be under-represented selection of invertebrate-rich areas for foraging use. in occurrence data and in the subsequent inferences. In This could be due to the availability of patchier food our study, resource selection value for general occur- such as fruits and crops, which may have been more rence showed a weaker correlation with roosting (an important than uniformly distributed invertebrates in elusive behavior) than with foraging (a conspicuous foraging decisions. Conversely, foraging birds might behavior). Literature on resource selection has rarely respond to invertebrate abundance at a larger scale addressed behavioral differences in habitat use, al- than our sampling frame; this needs to be examined though such differences are pervasive. Many animals in future studies. Courtship locations were relatively require multiple, contrasting habitats on a daily or sea- remote from settlements, as was also observed in the sonal basis and depending on age or sex (Law & Dick- case of displaying C. undulata in Morocco (Hingrat et man 1998). For instance, short-toed larks Calandrella al. 2008), reinforcing our understanding that rela- brachydactyla prefer areas dominated by short Salsola tively undisturbed habitats are essential for courtship plants for nesting and avoid areas do minated by activity. Although nesting sites appeared closer to cereals for feeding (Serrano & Astrain 2005). Black settlements, this was perhaps because nests were kites Milvus migrans prefer to nest in cliffs and trees at spatially clustered around an Air Force establish- rugged sites, and prefer to forage near water and ex- ment, whose effects differed from that of villages. tensive grasslands within 1 km of nest sites (Sergio et Although similar studies on other bustard species al. 2003). Other bustards require re latively tall sward have shown avoidance of infrastructure (Van Heezik for concealment of nesting fe males and short sward for & Seddon 1999, Sastre et al. 2009), we did not find a displaying males (Collar et al. 2017). strong negative response to roads and settlements for Finally, great Indian bustard usage at larger scales most behaviors. This finding could be an artefact of depended on the availability of useable and com - the sampling scale, as the gradient of anthropogenic plementary resources at finer scales. Availability of disturbances within the breeding habitat was proba- complementary microhabitats was a better indicator bly too narrow to elicit such behavioral responses. of patch quality, measured as bird density/usage, compared to availability of resources for general occurrence. This pattern could arise if birds examine 4.2. Complementary resources and and select patches with higher diversity of useable congruence between scales microhabitats in a top-down process. Proximity to complementary resources can be energetically ad - We found that great Indian bustards selected con- van tageous, as it reduces movement costs across trasting habitat characteristics for different activities larger areas to access spatially disjoint resources. that corroborated Rahmani’s (1989) qualitative de - This can attract more individuals and allow them to scriptions. Birds selected relatively taller sward for spend longer periods in an area. Multi-scale spe- nesting, shorter sward for courting and roosting, and cies−habitat analyses have commonly viewed habitat moderate-height sward for resting, thus partitioning selection as a similar hierarchical process with the spectrum of herbaceous vegetation into ‘sub- choices made progressively from coarser to finer niches.’ Such behavioral differences in microhabitat scales (Rettie & Messier 2000, but see Mayor et al. use has several consequences for resource selection 2009). Conversely, birds can directly select micro- Dutta & Jhala: Habitat selection by great Indian bustards 67

habitats, and the naturally occurring habitat hetero- areas to access their widely dispersed resources, and geneity may allow them to use those areas more this movement may increase fatal encounters with intensively where complementary resources are anthropogenic threats such as power-lines (Mahood available in close proximity. This would be a bottom- et al. 2018). Despite the ongoing advocacy, existing up process of habitat selection that has been largely power-lines are yet to be mitigated across several undermined in hierarchical habitat studies. thousand square kilometers of bustard landscapes, because of large financial costs and political inertia. However, relatively smaller areas encompassing crit- 4.3. Study limitations ical habitats can be secured on priority by burying power-lines. If these critical habitats are additionally This study has some shortcomings. Firstly, our managed to accommodate all behavioral require- inferences come from a diminishing population that ments and annual life history needs, then long- may raise concerns over maladaptive habitat choice, distance movements of the species can be partly re - causing the species to decline (Schlaepfer et al. stricted. Although this measure can reduce the risk of 2002). We cannot eliminate this possibility in the power-line mortality to some extent, it is not a suffi- absence of additional data on the species’ fitness cient measure to eliminate this threat. in different habitats. However, this scenario is un- While much great Indian bustard habitat is lost or likely since inferred resource characteristics largely degraded, some relatively small breeding areas matched with that of Rahmani (1989) when the pop- (5−15 km2) have been acquired by Forest Depart- ulation size was much larger and their habitat was ments for restoration (following Dutta et al. 2013). less altered. Further, our understanding of the spe- Managing these reserves by incorporating missing cies suggests that past hunting and power-line colli- resource units should benefit the species, according sions have contributed more to its decline than habi- to our findings. In a larger context, as wildlife habitat loss (Dutta et al. 2011, M. Uddin & S. Dutta, tats are shrinking, management of structural hetero- unpubl.). Secondly, RSFs based on direct observa- geneity leaves the possibility of increasing density/ tions have limitations, such as small samples of cryp- usage within remnant habitats for species that are tic behaviors and pseudo-replication. Some cryptic dependent on landscape complementation. Law & behaviors can be better captured through telemetry Dickman (1998) invoked the need for maintaining a (e.g. roosting and resting), while others are inher- patchwork of habitats and active management of ently rare/localized (courtship). However, models heterogeneity for conserving species that require ha- built on subsets of available locations showed similar bitat mosaics. Managing breeding habitats of great covariate effects, indicating that sample sizes were Indian bustards as mosaics of short and tall vegeta- likely adequate. The issue of pseudo-replication tion was recommended by Dutta et al. (2011) and could arise because of repeated sampling of same Collar et al. (2017). Greater usage of areas with con- individuals and their spatio-temporally auto-corre- trasting microhabitats, as shown by this study, bol- lated behaviors. We tried avoiding this problem by sters this proposition. sampling the area uniformly to capture several indi- Further, surrogacy is a popular management tool viduals, and restricting to only one ‘used’ location per that draws conservation focus towards species with behavior in an observation sequence. However, iconic, indicator or keystone values in the ecosystem issues such as individuals’ fidelity to nesting and (Caro & O’Doherty 1999). We propose that species other used sites could not be corrected. Thirdly, our de pending on heterogeneous habitats can be good sample size restricted us from exploring the effects of con servation proxies. Within the wide range of inter-annual variability (one sampling year experi- microhabitats used by great Indian bustards, other enced drought) and gender differences in habitat sympatric species are nested. The associated spiny- selection that need further examination. tailed lizard Saara hardwickii and Indian courser Cur sorius coromandelicus prefer short swards, whereas chinkara Gazella bennettii and quails Co - 4.4. Conservation implications tur nix spp. prefer moderate height swards (Dutta & Jhala 2014, S. Dutta pers. obs.). Targeting such spe- Despite the above issues, our finding that areas cies for conservation finds resonance in the well- with heterogeneous swards and diverse microhabi- established ‘habitat heterogeneity hypothesis’ that tats have greater use intensity carries strong man- postulates higher species diversity because of more agement implications. Bustards range over large niches in structurally complex environments (David- 68 Endang Species Res 45: 55–69, 2021

owitz & Rosenzweig 1998). A meta-analysis of this others (2004) Measuring global trends in the status of effect has found positive relationship between habi- biodiversity: Red List indices for birds. PLOS Biol 2:e383 Caro TM, O’Doherty G (1999) On the use of surrogate spe- tat heterogeneity and animal species diversity in cies in conservation biology. Conserv Biol 13: 805−814 85% of studies (Tews et al. 2004). Champion HG, Seth SK (1968) A revised survey of the forest Finally, in our study, behavioral ecology meets con- types of India. Government of India, New Delhi servation management through resource selection Choquenot D, Ruscoe WA (2003) Landscape complementation and food limitation of large herbivores: habitat- analysis. We show that the key to effectively manage related constraints on the foraging efficiency of wild critical habitats for a Critically Endangered bird lies pigs. J Anim Ecol 72: 14−26 in the ‘detail’ of microhabitat differences between Cody ML (1985) Habitat selection in birds. Academic Press, London be haviors. Statistical models of habitat use that Collar NJ, Baral HS, Batbayar N, Bhardwaj GS and others ignore these differences were poor indicators of (2017) Averting the extinction of bustards in Asia. Fork- habitat suitability and will be non-informative for tail 33: 1−26 species relying on habitat mosaics. At best, they can Davidowitz G, Rosenzweig ML (1998) The latitudinal gradient of species diversity among North American grass- predict occurrence of the species, but not why it hoppers (Acrididae) within a single habitat: a test of the occurs there, which is pivotal to its conservation. spatial heterogeneity hypothesis. J Biogeogr 25: 553−560 Dunning JB, Danielson BJ, Pulliam HR (1992) Ecological processes that affect populations in complex landscapes. Acknowledgements. The Wildlife Institute of India funded Oikos 65: 169−175 this research. We are grateful to the Director, Dean and Dutta S (2012) Ecology of the great Indian bustard (Ardeotis Research Coordinator of the Institute for administrative sup- nigriceps) in Kachchh, Gujarat with reference to report. We thank the Chief Wildlife Warden of Gujarat and the source selection in an agro-pastoral landscape. PhD dis- state Forest Department for granting research permission. sertation, Forest Research Institute, Dehradun We specially thank I. P. Bopanna and K. Maurya for their Dutta S (2018) Greener on neither side: socio-ecological cri- support during fieldwork, and B. Jethva for data sharing and sis of dry grasslands in India. In: Sreenivasan U, Velho N research inputs. We are grateful to I. Bhatti, L. S. Negi, (eds) Conservation from the margins. Orient Blackswan, R. Negi, I. Mundra, S. Maheshwari, L. Paradi and D. Paradi Hyderabad, p 198−231 for ﬁeld assistance. We also thank A. Rahmani for his input Dutta S, Jhala Y (2014) Planning agriculture based on land- in an earlier version of the manuscript, and S. Bisht for pre- use responses of threatened semiarid grassland species liminary proofreading. Finally, we thank 3 anonymous re - in India. Biol Conserv 175:129−139 viewers for their constructive comments to improve the Dutta S, Rahmani AR, Jhala YV (2011) Running out of time? manuscript. The great Indian bustard Ardeotis nigriceps — status, viability, and conservation strategies. Eur J Wildl Res 57: 615−625 LITERATURE CITED Dutta S, Rahmani A, Gautam P, Kasambe R, Narwade S, Narayan G, Jhala Y (2013) Guidelines for preparation of Akaike H (1974) A new look at the statistical model identifi- state action plan for resident bustards’ recovery pro- cation. IEEE Trans Automat Contr 19: 716−723 gramme. Ministry of Environment and Forests, Govern- Baasch DM, Tyre AJ, Millspaugh JJ, Hygnstrom SE, Ver- ment of India, New Delhi cauteren KC (2010) An evaluation of three statistical Graham MH (2003) Confronting multicollinearity in ecolog- methods used to model resource selection. Ecol Model ical multiple regression. Ecology 84: 2809−2815 221: 565−574 Gray TN, Chamnan H, Borey R, Collar NJ, Dolman PM BirdLife International (2018) Great Indian bustard Ardeotis (2007) Habitat preferences of a globally threatened bus- nigriceps. The IUCN Red List of Threatened Species tard provide support for community-based conservation 2018: e.T22691932A134188105. https:// dx. doi. org/10. 2305/ in Cambodia. Biol Conserv 138: 341−350 Gray TN, Chamnan H, Collar NJ, Dolman PM (2009) Sex- IUCN. UK.2018-2. RLTS. T22691932A134188105.en specific habitat use by a lekking bustard: conservation Boyce MS (2006) Scale for resource selection functions. implications for the critically endangered Bengal florican Divers Distrib 12: 269−276 (Houbaropsis bengalensis) in an intensifying agroeco - Boyce MS, McDonald LL (1999) Relating populations to system. Auk 126: 112−122 habitats using resource selection functions. Trends Ecol Green RE (1996) Factors affecting the population density of Evol 14: 268−272 the corncrake Crex crex in Britain and Ireland. J Appl Bretagnolle V, Villers A, Denonfoux L, Cornulier T, Ecol 33: 237−248 Inchausti P, Badenhausser I (2011) Rapid recovery of a Hilbert DW, Swift DM, Detling JK, Dyer MI (1981) Relative depleted population of little bustards Tetrax tetrax fol- growth rates and the grazing optimization hypothesis. lowing provision of alfalfa through an agri-environment Oecologia 51: 14−18 scheme. Ibis 153: 4−13 Hingrat Y, Saint Jalme M, Chalah T, Orhant N, Lacroix F Burnham KP, Anderson DR (2002) Model selection and (2008) Environmental and social constraints on breeding multimodel inference: a practical information-theoretic site selection. Does the exploded-lek and hotspot model approach, 2nd edn. Springer, New York, NY apply to the houbara bustard Chlamydotis undulata Burnham KP, Anderson DR, Laake JL (1980) Estimation of undulata? J Avian Biol 39: 393−404 density from line transect sampling of biological popula- Hovick TJ, Elmore RD, Fuhlendorf SD (2014) Structural tions. Wildl Monogr 72: 3−202 heterogeneity increases diversity of non-breeding grass- Butchart SHM, Stattersfield AJ, Bennun LA, Shutes SM and land birds. Ecosphere 5: 62 Dutta & Jhala: Habitat selection by great Indian bustards 69

IUCN (2018) The IUCN Red List of threatened species. Ver- Osborne PE, Alonso JC, Bryant RG (2001) Modelling land- sion 2018-2. www.iucnredlist.org (accessed 5 Feb 2021) scape-scale habitat use using GIS and remote sensing: a Jerrentrup JS, Wrage-Mönnig N, Röver KU, Isselstein J case study with great bustards. J Appl Ecol 38:458−471 (2014) Grazing intensity affects insect diversity via sward Quinn GP, Keough MJ (2002) Experimental design and data structure and heterogeneity in a long-term experiment. analysis for biologists. Cambridge University Press, J Appl Ecol 51: 968−977 Cambridge Johnsgard PA (1991) Birds of dry places: bustards, hemi - R Core Team (2013) R: a language and environment for sta- podes and sandgrouse. Oxford University Press, New tistical computing. R Foundation for Statistical Comput- York, NY ing, Vienna Johnson CJ, Nielsen SE, Merrill EH, McDonald TL, Boyce Rahmani AR (1989) Study of ecology of certain endangered MS (2006) Resource selection functions based on use− species of wildlife and their habitats: the great Indian availability data: theoretical motivation and evaluation bustard. Final report. Bombay Natural History Society, methods. J Wildl Manag 70:347−357 Bombay Jost L (2006) Entropy and diversity. Oikos 113: 363−375 Rettie WJ, Messier F (2000) Hierarchical habitat selection by Koshkin MA, Collar NJ, Dolman PM (2014) Do sheep affect woodland caribou: its relationship to limiting factors. distribution and habitat of Asian houbara Chlamydotis Ecography 23: 466−478 macqueenii? J Arid Environ 103: 53−62 Rushton SP, Ormerod SJ, Kerby G (2004) New paradigms for Koshkin M, Burnside RJ, Packman CE, Collar NJ, Dolman modelling species distributions? J Appl Ecol 41: 193−200 PM (2016) Effects of habitat and livestock on nest pro- Sastre P, Ponce C, Palacín C, Martín CA, Alonso JC (2009) ductivity of the Asian houbara Chlamydotis macqueenii Disturbances to great bustards (Otis tarda) in central in Bukhara Province, Uzbekistan. Eur J Wildl Res 62: Spain: human activities, bird responses and manage- 447−459 ment implications. Eur J Wildl Res 55: 425−432 Lane SJ, Alonso JC, Alonso JA, Naveso MA (1999) Seasonal Schlaepfer MA, Runge MC, Sherman PW (2002) Ecological changes in diet and diet selection of great bustards (Otis and evolutionary traps. Trends Ecol Evol 17:474−480 t. tarda) in north-west Spain. J Zool 247:201−214 Sergio F, Pedrini P, Marchesi L (2003) Adaptive selection of Lane SJ, Alonso JC, Martín CA (2001) Habitat preferences foraging and nesting habitat by black kites (Milvus of great bustard Otis tarda flocks in the arable steppes of migrans) and its implications for conservation: a multi- central Spain: Are potentially suitable areas unoccupied? scale approach. Biol Conserv 112: 351−362 J Appl Ecol 38: 193−203 Serrano D, Astrain C (2005) Microhabitat use and segrega- Lavee D (1988) Why is the houbara Chlamydotis undulata tion of two sibling species of Calandrella larks during the macqueenii still an endangered species in Israel? Biol breeding season: conservation and management strate- Conserv 45:47−54 gies. Biol Conserv 125: 391−397 Law BS, Dickman CR (1998) The use of habitat mosaics by Shoemaker KT, Heffelfinger LJ, Jackson NJ, Blum ME, terrestrial vertebrate fauna: implications for conserva- Wasley T, Stewart KM (2018) A machine-learning tion and management. Biodivers Conserv 7: 323−333 approach for extending classical wildlife resource selec- MacDonald AJ, Mayer P, Armstrong HM, Immirzi, P, tion analyses. Ecol Evol 8:3556−3569 Reynolds P (1998) A guide to upland habitats. Surveying SoIB (2020) State of India’s Birds, 2020: range, trends and Land Management Impacts. Scottish Natural Heritage, conservation status. The SoIB Partnership, India Battleby Stephens DW, Krebs JR (1986) Foraging theory. Princeton Magana M, Alonso JC, Martin CA, Bautista LM, Martin B University Press, Princeton, NJ (2010) Nest-site selection by great bustards Otis tarda Tews J, Brose U, Grimm V, Tielbörger K, Wichmann MC, suggests a trade-off between concealment and visibility. Schwager M, Jeltsch F (2004) Animal species diversity Ibis 152: 77−89 driven by habitat heterogeneity/diversity: the impor- Mahood SP, Silva JP, Dolman PM, Burnside RJ (2018) Pro- tance of keystone structures. J Biogeogr 31: 79−92 posed power transmission lines in Cambodia constitute a Thomas L, Buckland ST, Rexstad EA, Laake JL and others significant new threat to the largest population of the (2010) Distance software: design and analysis of distance Critically Endangered Bengal florican Houbaropsis sampling surveys for estimating population size. J Appl bengalensis. Oryx 52:147−155 Ecol 47: 5−14 Manly BF, McDonald LL, Thomas DL, McDonald TL, Erick- Tilman D, May RM, Lehman CL, Nowak MA (1994) Habitat son WP (2002) Introduction to resource selection studies. destruction and the extinction debt. Nature 371: 65−66 Resource selection by animals: statistical design and Tourenq C, Combreau O, Lawrence M, Pole SB and others analysis for field studies, Kluwer Academic Publishers, (2005) Alarming houbara bustard population trends in New York, NY Asia. Biol Conserv 121: 1−8 Mayor SJ, Schneider DC, Schaefer JA, Mahoney SP (2009) Van Heezik YV, Seddon PJ (1999) Seasonal changes in Habitat selection at multiple scales. Ecoscience 16: habitat use by houbara bustards Chlamydotis [undulata] 238−247 macqueenii in northern Saudi Arabia. Ibis 141: 208−215 McCullagh P, Nelder JA (1989) Generalized linear models, Vittinghoff E, McCulloch CE (2007) Relaxing the rule of ten 2nd edn. Monogr Stat Appl Prob 37. Chapman & Hall/ events per variable in logistic and Cox regression. Am J CRC, Boca Raton, FL Epidemiol 165: 710−718 Nad’o L, Kaňuch P (2018) Why sampling ratio matters: logis- White RP, Murray S, Rohweder M, Prince SD, Thompson tic regression and studies of habitat use. PLOS ONE 13: KM (2000) Grassland ecosystems. World Resources Insti- e0200742 tute, Washington, DC O’Brien M, Green RE, Wilson J (2006) Partial recovery of the Wolff A, Paul JP, Martin JL, Bretagnolle V (2001) The bene- population of corncrakes Crex crex in Britain, 1993− fits of extensive agriculture to birds: the case of the little 2004. Bird Study 53: 213−224 bustard. J Appl Ecol 38:963−975

Editorial responsibility: Michael Reed, Submitted: September 15, 2020 Medford, Massachusetts, USA Accepted: March 16, 2021 Reviewed by: 3 anonymous referees Proofs received from author(s): May 19, 2021